Radiation Protection Dosimetry (2007), Vol. 126, No. 1 4, pp. 229 233 Advance Access publication 11 May 2007 doi:10.1093/rpd/ncm047 CHARACTERIZATION AND UTILIZATION OF A BONNER SPHERE SET BASED ON GOLD ACTIVATION FOILS D. J. Thomas*, N. P. Hawkes, L. N. Jones, P. Kolkowski and N. J. Roberts Neutron Metrology Group, DQL, National Physical Laboratory, Teddington, Middlesex, TW11 0LW, UK Bonner sphere (BS) sets which use activation foils as the central thermal neutron sensor have advantages over active BS systems in certain environments, for example, pulsed fields, or fields with high photon components. In such environments, they may be the only type of neutron spectrometer which can be used. This paper describes work, using both measurements and calculations, to validate the response functions for a BS set based on gold activation foils. As an illustration of the use of such a system, a measurement is described of the contaminant neutron spectrum in the treatment room of a 21 MV hospital linear accelerator providing photon beams for radiotherapy. INTRODUCTION A measurement of the neutron spectrum in a mixed n/g field where the radiation is pulsed presents certain problems. This is particularly true if a measurement is required over an extensive range of neutron energies, from thermal to 20 MeV or higher, for example. Such measurements are often made with a Bonner sphere (BS) set using an activation material, for example gold or indium, as the central sensor. Gold is the most commonly used material, and this work describes the characterization of a BS set based on gold foils, and its use to measure the contaminant neutron spectrum in the vicinity of a hospital electron LinAc used for cancer therapy where the radiation field is pulsed, has an high photon component, and the neutron spectrum extends over a wide energy range. The paper describes a continuation of work with gold-foil-based BSs at NPL (1,2). It presents further validation of the response functions of the sphere set thus enabling other users to build and use BSs of the same design without the need for extensive characterization of response functions. THE GOLD-FOIL-BASED BONNER SPHERE SET The sphere set and how it is used The gold-foil-based BS set currently used at NPL consists of 10 spheres, of high density polyethylene (0.96gcm 23 ), with diameters of 2 00,2.5 00,3 00,3.5 00,4 00, 5 00,6 00,8 00,10 00 and 12 00. (BSs are usually made to exact inch sizes, and the diameter in inches acts as a label for the sphere.) A small slot at the centre of each sphere accommodates the gold foil. Each foil is 4 cm 2 in area, 0.05 mm thick and weighs about 360 mg. *Corresponding author: david.thomas@npl.co.uk Activity is induced in the foils via 197 Au(n,g) 198 Au reactions. The induced activity is usually low, and so is determined by b-counting in a low-background 4p proportional counter which provides a counting efficiency of roughly 40% for foils of this thickness. b-counting efficiencies are determined by irradiating the foils in a reactor to sufficiently high activities that 4pb-g coincidence counting can be performed from which the efficiency can be determined (2). Values for the b-counting efficiencies of thin foils depend on the distribution of the activity with depth in the foils, and vary somewhat depending on the field in which they are irradiated. The neutron field at the centre of a BS is reasonably well thermalized, and b-counting efficiencies determined in a reactor are thus generally applicable. However, small corrections do need to be made for the small spheres where the field at the centre may be less well thermalized. Correction factors are available from measurements made in a high-energy neutron field at an intensity which was sufficiently high to allow 4pb-g counting of the foils. These corrections are small, with a maximum value of just over 2% for the 2 00 sphere. In high-intensity high-energy gamma-ray fields, gold can be activated via the 197 Au(g,n) reaction producing 196 Au. This isotope has a half-life of 6.2 d, and can be separated from the 2.7 d half-life 198 Au by measuring the count rate over several days and analysing the data assuming two components with different half lives. Derivation of the response functions Because the efficiency of this BS set is low, it is difficult to measure response functions using monoenergetic neutron calibration fields. The data for this set were therefore originally (2) derived primarily from calculations with the ANISN transport code (3). This code has the advantage that, when run in adjoint David Thomas, National Physical Laboratory. # Crown copyright 2007. Reproduced with the permission of the Controller of Her Majesty s Stationery Office.
mode, it can calculate complete response functions in a single, relatively short, computer run. From experience deriving BS response functions for spheres with a 3 He counter as the central sensor ANISN was expected to give good response function shapes for the gold foil BSs, particularly at high and intermediate energies, although some normalization is likely to be required (4), but the predictions are expected to be higher than measurements in the thermal region (5). Thus, the ANISN calculations were normalized using results from measurements with a high-output 252 Cf source. Furthermore, the responses in the thermal region were modified to agree with a limited number of MCNP (6) calculations. (The MCNP calculations were limited in number because they are performed at point energies, and at the time the response functions were originally derived, circa 1995, numerous lengthy MCNP runs took a prohibitively large amount of computer time.) For BS unfolding, the response functions are required in group format and the ANISN-based results, referred to as the Mk 1 response functions, are shown for two spheres as histograms in Figure 1. With present-day computing speeds, the extensive set of calculations required to calculate response functions with MCNP can readily be performed, D. J. THOMAS ET AL. and a full set of calculations was made for all the spheres for the mid point energies of the 52 bins used in the Mk 1 response functions. Neutrons were incident isotropically on the BSs. A type 4 tally was used, with the appropriate multiplier, to give the activity of the gold foil at saturation. These data are shown in Figure 1 for the 2 00 and 6 00 spheres. Except for some details of the shape in the region of 1 ev for the small spheres, the agreement with the Mk 1 response function is very good. To check the absolute values of the MCNP calculations, the response to a 252 Cf source was calculated, and in Figure 2(a) these values are compared with the experiment. The measurements were performed in the low-scatter facility at NPL. MCNP calculations were performed using the 252 Cf spectrum recommended by ISO (7). Also shown in Figure 2(a) is the values for the original, un-normalized, ANISN calculations derived by folding the calculated response functions with the spectrum. The error bars are the experimental standard uncertainties, combined in quadrature with the statistical uncertainties in the case of the Monte Carlo codes. Figure 1. Comparison of latest MCNP calculations with response functions reported in ref. 2. Figure 2. Comparison of experimental and calculated results for a 252 Cf source and for a thermal neutron beam. 230
From the evidence of the 252 Cf measurements, it is clear that MCNP is better at getting the values for the responses correct on an absolute scale, but even for this code there are some differences between experiment and calculation. Interestingly, the 252 Cf measurements indicate that both ANISN and MCNP overestimate the response for the smaller spheres, albeit by different amounts. The underestimation by both codes of the response of the 8 00 sphere suggests possible differences between the parameters used in the calculations and those applying to the actual sphere, a void in the polyethylene for example. Since the 252 Cf experimental data were used to normalize the Mk 1 response functions, the value of the experimentally measured response divided by the response predicted by folding the Mk 1 response functions with the 252 Cf spectrum is essentially unity for all spheres. The availability of reasonable thermal neutron fluence rates ( 4 10 4 cm 22 s 21 )atnpl (8) meant that the response functions for the smaller spheres could be checked at thermal energies. Measurements were made in a thermal column beam, for spheres with diameters up to 6 00, with the spheres both bare and under 1 mm of cadmium. In this way the responses to sub-cadmium-cut-off neutrons were determined (5). The spectrum of these neutrons, at the measurement position in the beam, consists of a Maxwellian peak, at a temperature of 323 K, and a small 1/E contribution extending up to the cadmium cut-off energy of 0.5 ev. The experimental thermal responses are compared with MCNP calculations in Figure 2(b) by plotting the ratio of experiment/calculation. Although there is agreement for the 4 00,5 00 and 6 00 spheres, MCNP overestimates the responses of the smaller spheres, and the overestimation increases as the sphere size decreases reaching a value of 15% for the 2 00 sphere. Discrepancies of this order are much larger than any found when comparing 252 Cf experiments and MCNP. To investigate these discrepancies thermal response calculations were also performed with the MCBEND code (9). This is a general purpose 3-D Monte Carlo code which can calculate neutron, gamma and charged particle transport in sub-critical systems. It is one of a suite of programs available from the ANSWERS software service (9). The data are plotted in Figure 2(b) and it can be seen that there is very good agreement between the results of the two Monte Carlo codes. There is no obvious reason why these two codes should predict accurately the thermal responses of the 4 00 to 6 00 spheres, but not the responses for smaller spheres. The gold foils are 2.3 cm in diameter which means that the foil-plus-sphere combination is not truly spherically symmetric, and the responses of the spheres may vary with angle of incidence. This effect is likely to be most pronounced for the small spheres THE GOLD-FOIL-BASED BONNER SPHERE SET 231 where the foil radius is a significant fraction of the sphere radius. A set of calculations was performed with MCNP, of the responses for all the spheres for neutrons incident normal to the foil, and parallel to the foil, and these were compared with the calculations for isotropic incidence. The results are shown in Figure 3 for the 2 00 and 5 00 spheres. For the smaller spheres at low energies MCNP predicts a definite anisotropy although this effect becomes increasingly unimportant as sphere size and neutron energy increase. To check these calculations, a measurement was performed for the 2.5 00 sphere in the NPL thermal column with the foil both normal and parallel to the beam. The measured value for parallel/normal responses was 1.086 + 0.019 which compared well with the calculated value of 1.076 + 0.014. Although anisotropy needs to be considered for the smaller spheres at thermal energies, it cannot explain the discrepancy between experiment and calculation for the four smallest spheres since the experiments and calculations were performed for the Figure 3. Responses for normal and parallel neutron incidence relative to the plane of the gold foil.
same orientation of the foil relative to the neutron beam. In the absence of an explanation for the discrepancies allowance is made when deriving spectra by increasing the uncertainties for the low-energy responses of the smallest spheres. The effect on a derived spectrum, and in particular on the dose derived from such a spectrum, is small. MEASUREMENTS AT WALSGRAVE HOSPITAL A neutron spectrum was measured at the Precise Electa Magnetron accelerator at Walsgrave Hospital, Coventry, UK. At this facility the electron beam is accelerated to 21 MeV, but an aluminium hardening plug in the primary collimator filters the lower energies from the target, giving the beam a depth dose profile corresponding to 25 MV. The treatment dose rate was 4 Gy min 21, with a field size of 10 10 cm at the isocentre. Measurements were performed on the treatment couch at 1 m from the axis of the photon beam. The spheres were irradiated sequentially for a time corresponding to a treatment dose of 20 Gy. Foils were returned to NPL to determine the saturated 198 Au activities which ranged from 66 to 460 Bq mg 21. The 196 Au activities were negligible. The neutron spectrum was unfolded using the STAY SL code (10) and the result is shown in Figure 4 where it is compared with an earlier measurement (2) for the corresponding site on the treatment couch of a 15 MV Varian Clinac 2100C accelerator at St Bartholomew s Hospital, London. It is clear that despite the different electron beam energies, the spectra at the two facilities are similar in shape when normalized to unit fluence. As expected the energy of the peak is a little higher at Walsgrave, the overall mean energy being 318 kev Figure 4. Comparison of fluence spectra measured at 1 m from the beam axis at two different hospitals. D. J. THOMAS ET AL. 232 Table 1. Comparison of the neutron fluence and dose values at Walsgrave and St Bartholomew s hospitals. Hospital Total neutron fluence (cm 22 Gy 21 ) Total ambient dose equivalent (msv Gy 21 ) Total dose (ICRU muscle) (mgy Gy 21 ) Walsgrave 18.4 10 6 2.86 0.16 St Barts 2.92 10 6 0.37 0.02 Ratio 6.3 7.8 8 compared to 216 kev at St Bartholomew s. At the higher energy facility, the neutron fluence and dose rate are, however, significantly higher as shown in Table 1. Nevertheless, the neutron dose is still smaller than 0.02% of the dose in the beam, the value recommended by the IEC (11) as a limit for the average neutron dose in the patient plane. CONCLUSIONS The response functions of the NPL gold foil BS set have been further validated. Although there remain some uncertainties for the small spheres at thermal energies, these do not prevent the system being used to provide valuable information about the spectra of neutrons around medical accelerators. ACKNOWLEDGEMENTS The authors would like to acknowledge the help provided by John Mills and Roger Aukett at Walsgrave Hospital. REFERENCES 1. Axton, E. J. and Bardell, A. G. Neutron production from electron accelerators used for medical purposes. NBS Special Publication, 554, 109 123 (1979). 2. Thomas, D. J., Bardell, A. G. and Macaulay, E. M. Characterisation of a gold foil based Bonner sphere set and measurements of neutron spectra at a medical accelerator. Nucl. Instrum. Meth Phys. Res. A 476, 31 35 (2002). 3. Engle, W. W. A user manual for ANISN. ORNL Report K-1693 (1967). 4. Alevra, A. V. and Thomas, D. J. Neutron spectrometry in mixed fields: multisphere spectrometers. Radiat. Prot. Dosim. 107, 37 72 (2003). 5. Thomas, D. J., Alevra, A. V., Hunt, J. B. and Schraube, H. Experimental determination of the response of four Bonner sphere sets to thermal neutrons. Radiat. Prot. Dosim. 54, 25 31 (1994). 6. Briesmeister, J. F. Ed. MCNP TM, Version 4C LA- 13709-M (2000). 7. International Organization for Standardisation, Reference neutron radiations: calibration of area and personal
dosimeters and determination of their response as a function of energy and angle of incidence. International standard ISO 8529-3 (Geneva: ISO) (1998). 8. Ryves, T. B. and Paul, E. B. The construction and calibration of a standard thermal neutron flux facility at the National Physical Laboratory. J. Nucl. Energ. 22, 759 775 (1968). 9. Cowan, P., Shuttleworth, E., Bird, A. and Cooper, A. The Launch of MCBEND 10. In: 10th International Conference on Radiation Shielding (ICRS-10) and 13th Topical Meeting on Radiation Protection and Shielding THE GOLD-FOIL-BASED BONNER SPHERE SET (RPS-2004), Funchal, Madeira, Portugal (May 2004), ANSWERS Service, www.sercoassurance.com/ ANSWERS/. 10. Perey, F. G. Least squares dosimetry unfolding: the program STAY SL. Oak Ridge National Laboratory Report ORNL/TM-6062, ENDF-254 (1977). 11. International Electrotechnical Commission. Medical electrical equipment Part 2.1 Specification for medical electron accelerators in the range 1 ev to 50 MeV. International Standard IEC 60601-2-1 (Geneva: ISO) (1998). 233